Sarah McConnell & Sunil Rao 19 min

Scale Faster with Tribble AI Autonomous Agents


Sunil Rao, Founder & CEO at Tribble, shows us how Autonomous AI Agents can integrate into your systems and learn from your data.



0:00

[MUSIC]

0:04

Hello everyone and welcome to Go to Market AI,

0:07

the future of your Go to Market tech stack.

0:09

I'm your host Sarah McConnell.

0:11

These days it seems like every product has AI,

0:14

but on this show we want to go a level deeper,

0:16

so you can see firsthand how businesses are

0:19

actually applying AI to solve your use cases.

0:22

We're going deep into those use cases and showing you

0:24

live demos of the latest and greatest in AI technology.

0:27

Today I'm joined by Sunil Rao, founder and CEO at Triple.

0:31

Sunil, welcome. Thank you for joining us on the show today.

0:34

>> Thanks for having me, Sarah.

0:36

>> All right. I'm really excited.

0:38

I will admit that I have not seen a demo of

0:40

Triple before, but before we jump into that, I'd love to know.

0:43

Who is Triple? What do you guys do and who are you helping in the market?

0:47

>> Yeah. Once again, thanks for having me.

0:50

Super excited to talk to you guys today about what we do.

0:53

Triple is all about Go to Market automation.

0:57

A little bit of the background of the founder and CEO and my co-founder of Ray,

1:02

we're both ex-Silsforce employees.

1:04

In our experience scaling teams at Salesforce,

1:08

one of the things that we came across quite a bit was this synergy

1:13

across Go to Market teams,

1:14

thinking through how we bring together these new products that we're

1:18

launching at the company and how we bring them to scale and bring them to

1:21

customers.

1:22

We both shared many roles at the company.

1:25

But I think suffice to say,

1:27

one of the biggest challenges we came across was scaling deep product knowledge

1:32

and deep industry knowledge across the presales and Go to Market teams in the

1:37

company.

1:38

If we think about who Triple is meant to help,

1:41

it's meant to help B2B companies scale their Go to Market teams.

1:45

I think if you think about how we've built the product and how we're thinking

1:51

about

1:51

what it means for our customers,

1:53

we're seeing a lot of proliferation of SaaS technology out there.

1:56

I mean, this is no surprise to anyone.

1:58

People have SaaS fatigue, I think,

2:00

out there if you talk a lot of customers.

2:02

And we're all about kissing these apps goodbye.

2:06

We genuinely think that we are now in a new phase of how software will be

2:12

developed.

2:12

Just the other day, Bill Gates put this blog post out talking about how AI

2:16

agents are the future.

2:18

And I know you're going to hear a lot of that terminology coming up

2:20

and people are going to start talking about it now.

2:22

But we really believe that from day one.

2:25

And that's how we've been building our product.

2:26

It's about these agents that go in.

2:29

They integrate with the systems you already have.

2:32

They learn from your data and they take action within the tools you already

2:36

have.

2:36

I think that's a key point.

2:38

Reuse what you have and maximize the utility of those things.

2:42

That's amazing.

2:43

And I'm so excited to now move into the demo

2:45

because the fact that you guys are meant for Go to Market teams,

2:48

and this is Go to Market AI,

2:49

I can't think of a better demo to get on the shuffle.

2:52

So that being said, Sunil, I would love to jump into the demo

2:55

and just see what tribals all about and see your AI functionality in action.

2:59

Excellent.

3:00

So right before I jump in, just some context setting,

3:03

you know, we're going to focus on one out of the multiple agents that we have.

3:08

We'll focus on the questionnaire RFP agent.

3:11

But in how we built out the stack, really, there's three, right?

3:15

There's this questionnaire RFP agent.

3:16

There's a sales engineering agent.

3:18

We'll touch on that just briefly during the demo today.

3:21

And we've also got this content agent for marketing,

3:23

which we will talk about, but not be shillion.

3:25

Often.

3:26

Look at that.

3:27

The beautiful document that is the RFP, beautifully colored

3:31

and in all its full glory that every seller out there flinched a little bit.

3:36

They saw it and just kind of, "Ugh."

3:38

And look, you know, I've had much of these.

3:41

I used to be a sales engineer many years ago.

3:44

And these things, you know, they take a lot of time

3:47

and these things are still rampant,

3:48

whether they're in the form of actual RFPs or they are infosex,

3:52

where, you know, you now have to provide a lot of the security

3:55

and compliance information back to your customers

3:57

about what your software can, can't do, etc.

4:00

So I'm going to play the role of Daniel, who's actually one of our engineers.

4:05

And I'm going to quickly walk through what Daniel does when one of these comes

4:10

in at trouble.

4:12

So, you know, he kind of opens this up.

4:14

And typically, you know, you're working out of a Google sheet

4:17

or you're working in some weird portal where, you know,

4:20

the structure of these questions reveals itself as you click this two-zero

4:24

adventure navigation.

4:26

But regardless, the point is let's read Daniel where he lives, which is in the

4:30

browser.

4:31

And as a result, the first modality for this RFP questionnaire agent

4:36

is a Chrome extension inside the browser.

4:38

So here I've clicked this button on the top right

4:40

and it's popped open the triple Chrome extension.

4:42

And very quickly, you can see Daniel can see

4:45

all of the different questionnaires or infosecs or RPs

4:48

he's already started in his answer.

4:50

And if I scroll down over here, I've already got one for info

4:54

for the infosec for OpusTech.

4:56

When I pop that open, you can see some of the questions

4:59

that are in this document on the left that have already been sent in by Dan.

5:03

And the intention here really is, hey, I've got all these questions.

5:08

I now typically would have to go through these

5:10

and maybe use some legacy software,

5:12

which has a Q&A data bank and a whole bunch of project management stuff.

5:16

So can I just get straight to the answer?

5:18

Like, can't I just have this generate answers?

5:21

And that's really what triple isn't intended to do in this scenario.

5:24

So I'm going to very quickly select all these questions here on the left side.

5:29

And when I come over here on the right in the panel, I can go in

5:32

and I can add all these questions when I paste it.

5:35

Triple immediately recognizes that there are 18 of them.

5:38

And then I can say, look, all these questions that I just added to the system

5:42

actually, they're all security related questions.

5:44

So I'm going to assign them and apply it and immediately organize this.

5:50

And then I click on generate answers and triples off to the races.

5:54

The intention here is Daniel doesn't have to spend the time looking for answers

5:59

Triple will generate them.

6:00

And as they come in, Daniel can now take a look at what's come in

6:04

and dig in to whether or not the answer is acceptable.

6:06

So let me click over here on this first one, which is already pre-generated.

6:11

And Daniel can quickly see, hey, there's some detailed documentation that was

6:17

referenced.

6:17

You can actually even see the reference numbers of the documentation here.

6:20

Hello.

6:21

As he reads this, if you want to understand a little bit more about where it

6:24

came from,

6:24

you can click on sources.

6:25

This is actually all of our internal stock to compliance talks, right?

6:28

Stuff that Daniel and T.

6:30

don't like digging through in order to find an answer.

6:32

But you can very quickly see all the different sources that were referenced

6:38

and how they were pulled into the answer.

6:41

And the interesting thing here, Sarah, is when you sometimes answer these

6:44

questions,

6:45

you might not want to generate a new answer.

6:47

You might actually see something in an existing doc that was good enough as is.

6:50

And in that case, you can just quickly use that answer verbatim so that you're

6:54

not creating

6:55

something new or deviating away from the general acceptable answer that's it.

6:59

That's really amazing.

7:01

I love a theme that we've seen a little bit on this show that I don't think

7:05

actually we've

7:05

seen enough of in all the products is the sources.

7:08

And I love that you've built into the UX here that I can just see where that

7:12

came from.

7:13

So I trust what's being generated.

7:15

I know what sourced it.

7:16

I can go in and to your point, just select something or use that as a

7:21

foundation.

7:21

I also could absolutely see a use case where you find some outdated documents

7:25

that you didn't

7:25

even know were outdated because you see it's pulling and you're like,

7:28

oh, I got to go change this in our systems because I forgot that this is our,

7:31

you know, our info sec from last year, whatever it might be, our stock to from

7:34

last year.

7:35

So this is really, really cool.

7:37

I love that you went there, Sarah, because that is precisely one of the first

7:40

things

7:40

that one of our customers brought up, which is like, hey, this is a look right.

7:43

It's like, well, when you debug further, you realize that your documentation

7:47

wasn't updated.

7:48

And, you know, I've been part of good market teams for a long time, even at

7:52

Salesforce.

7:53

I recall how many times documents go stale, right?

7:56

So it's like, this gives you the ability to also crowdsource the constant

8:00

update without

8:00

having to do these batch updates at the beginning of the year that take a lot

8:03

of time

8:04

as you kind of hunt and pack these things down.

8:06

So, you know, that was one question for Daniel.

8:09

He's got a bunch more to go.

8:10

And as he goes through these, you know, you can kind of go in.

8:13

And one of the things you can also do is mark them for review.

8:15

So light workflow instead of having a heavy project management base workflow,

8:20

really so that it's focused on getting the answers out.

8:23

You know, one of the other things we've heard from our customers is most people

8:27

have an SLA and turnaround time of like a week when they get one of these.

8:30

And it's difficult to turn around if you don't have some assistance.

8:34

So very quickly going down here, let's take a look at this question.

8:38

Opus technology prioritized data security compliance.

8:42

Daniel takes a look at this and he says, wait a minute,

8:45

HIPAA compliance, we're not HIPAA compliant.

8:47

Maybe that sets off an alarm.

8:49

Now that this could be for any reason, either Daniel is not up to date or the

8:54

documentation is not.

8:55

But in this case, you know, it's probably Daniel that's not up to date.

8:57

So he goes in and says, look, I'm going to, I'm going to provide it at eight

9:01

here.

9:01

Because I think that this is actually not correct.

9:06

So I'm going to, I'm going to come in here and say, although

9:09

we are not HIPAA compliant.

9:16

And then this route, this part over here.

9:19

Okay, great.

9:21

That looks great.

9:21

So now it's not talking about us being HIPAA compliant.

9:23

I won't get in trouble.

9:24

And then he hits say.

9:27

So now you might not necessarily want people to be able to make these kind of

9:32

modifications

9:33

without having some sort of a checking mechanism in the background.

9:36

Like, yeah, just overrated something in the back.

9:38

And listening to our customers, you know, this is one of the key things where

9:42

we really didn't

9:43

want to rebuild workflow into the tool.

9:45

We wanted this to be in a tool that they already live and work in.

9:47

So with that, if I quickly slip this change over now, and now I'm going to play

9:53

myself,

9:54

Sunil, I'm the approver of Daniel's request.

9:57

Oh, and look, that's a wonderful offsite from last week, the team having dinner

10:02

But then here on the left side, I see a Q&A approval notification.

10:06

And when I click on that, I see that that questionnaire triggered an approval

10:11

directly

10:12

in Slack using the workflow capability.

10:14

So once again, it's that questionnaire agent tied into the tools that you work

10:18

in,

10:18

taking action, asking me for a decision.

10:21

And then I take a look at this as a reviewer and I'm like, wait a minute, that

10:25

doesn't sound right.

10:26

We are HIPAA compliant.

10:27

Why is Daniel saying this?

10:29

So what do I do?

10:30

Well, the beauty of it is now with the sales engineer agent,

10:34

which is another manifestation of the RFP agent, I can add mention,

10:37

triple directly here and say, are we HIPAA compliant?

10:42

You know, I'm not going to call out Daniel by name, but let's see what's

10:46

triple and understand why Daniel might be suggesting that change.

10:52

I'm going to quickly click on reply over here and close the extension so we see

10:56

the whole screen.

10:57

Like, oh, okay, well, that's pretty clear.

10:59

We've become HIPAA compliant a week ago.

11:01

So this is very clearly documented and I have a reference over here.

11:05

The beauty of this is this agent is not just connected to data sources,

11:10

it's also connected to systems.

11:12

So now I could do something like at triple how, you know, Daniel's talking to

11:18

Opus, was it?

11:19

So this is for Opus, heck, why do they need HIPAA?

11:28

And this could be any kind of a question related to the data about the customer

11:32

that may be

11:32

locked up in another transactional system.

11:34

In this case, you know, it's going to pull up that specific account record over

11:38

here

11:39

that's sitting in your external system and it's telling you the reason why,

11:43

and this in this case, it actually took it from a notes file that was attached

11:46

to that record.

11:47

But what's happening here is the sales engineer agent has access to systems as

11:50

well, not just

11:51

the data that we fed it.

11:52

And it can now help debug and look at the system and take action within it.

11:57

Before I decide, hey, actually, Daniel, I reviewed your request.

12:02

I'm going to click on this decision here, say declined.

12:06

We are already HIPAA compliant plus C was like Opus needs this, whatever that

12:16

case may be.

12:17

So that's in a nutshell, the tie in of how the questionnaire agent works

12:24

alongside

12:25

the digital SE that's in Slack and all these things come together for our

12:31

customers.

12:32

So, both of it?

12:34

So cool. So being at a mid-sized company, I still see, I'm at some of it,

12:39

I joined when we were very much smaller.

12:41

So I'm still in a lot of Slack channels that I don't necessarily know anything

12:45

about.

12:45

But a lot of times I see when we get RFPs, RSEs, we're fantastic, asking a lot

12:51

of these security

12:52

questions. And every once in a while, if I go into the threads, you can see how

12:58

there's confusion.

12:59

Like stuff changes rapidly, data security, they live in a lot of different

13:03

systems.

13:04

And you'll even see people kind of going back and forth in Slack about like,

13:08

is this the right answer? Are we actually compliant here?

13:10

And it just almost reminded me of like, you can just ask it like Google, like,

13:15

hey,

13:15

why do we need this? And it's going instead of that SE having to then go into

13:19

your CRM data or into

13:21

wherever your repository for your documents are and try to hunt this down,

13:24

which is going to take a ton of time.

13:26

That was so fast. And he showed that in just a couple of minutes.

13:29

That's incredible. And I know our SEDM, no one loves an RFP. They do take a lot

13:34

of time.

13:35

And I have to imagine this is just speeds up that process immensely.

13:40

Yeah, no, sir. That's spot on. I think, you know, the customers that we have

13:44

right now,

13:44

the business cases are very clear cut, right? It's actual amount of hours saved

13:49

and time back.

13:50

So those folks can focus on the more strategic conversations in the deals where

13:54

SE should be frankly spending their time not hunting down knowledge.

13:57

And as your team grows, that stuff starts to get really spread out.

14:01

There's a lot of people information doesn't get passed around as easily. We're

14:05

all virtual.

14:05

So yeah, I see a ton of use cases here.

14:08

Yeah. And the last thing I'll say, Sarah, is, you know, the interesting thing

14:16

is we're at this

14:17

precipice where all of a sudden what can happen with these tools is broaching

14:23

what we're only was

14:24

only restricted to the domain of humans, right? Like people could only like you

14:28

would get hired

14:29

into a new role at a company. You get this Google Drive of training docs put in

14:33

front of you.

14:34

You get three months to get onboarded and you get a bunch of SaaS licenses and

14:38

you got to figure

14:38

out how to weave all this stuff together. Well, what we're doing is very much

14:42

looking at

14:43

parts of that stack and saying SaaS licenses to different systems. Yeah, we can

14:46

connect to those.

14:47

We'll connect to that Google Drive or whatever the input training content is.

14:51

We'll train the tool on that information and allow it to take actions and

14:55

assistance.

14:55

And I think we're going to see a lot more of that paradigm take shape.

14:59

Yeah. Well, that's amazing that you guys are on the forefront of this. And with

15:02

that being said,

15:03

I would love to transition us into our lightning round Q&A section. If that's

15:07

okay with you.

15:07

Yeah. The first question here being how long have you been building AI into

15:13

triple?

15:13

So I would say like, can I give you an answer, which is negative in time?

15:18

Because before we even started triple, I think as early back as GPT-2,

15:24

when one of these first language models became a little bit more prevalent at

15:28

OpenAI,

15:28

everybody's something, I had already started playing with it and looking at use

15:33

cases in terms

15:34

of how we can help create content and wired into the workflows within GoTo

15:39

Market. But joking aside,

15:41

since day one, Sarah, I think that's one of the most amazing things of that

15:46

this journey has been

15:47

how we as a team are baking all this technology into everything that we do from

15:53

day one.

15:54

And building an AI first company is really exciting because I feel like a kid

15:58

in a candy store

15:59

every day, there's something new that we can incorporate to make the product

16:02

better for our

16:03

customers. I love asking that question to any company that has .ai as their

16:07

domain,

16:08

which if you're listening to this and you want to check out triple, triple.ai

16:11

is the domain,

16:12

because I know the answer is going to be, this has been the foundation of our

16:15

product since day one.

16:16

So love that. Now, what you showed in the demo today, is it generally available

16:21

now for anyone

16:22

if they were to go purchase triple today or tomorrow? It is. It is generally

16:26

available today.

16:27

You can download that Chrome extension, but it won't do you until you have an

16:30

account.

16:31

And one of the things that we're doing with all of our customers is spending a

16:37

lot of time front

16:38

thinking through how we load their information in a trusted way and integrate

16:43

with their systems

16:44

in order to make what we just showed you on the screen.

16:46

Amazing. And speaking of customers, who are some of the current customers that

16:51

are benefiting

16:51

from Triple's AI functionality? We're very fortunate to have some awesome

16:57

customers,

16:57

even since the earliest of days. We're working closely with the teams over at

17:01

UICath

17:02

and over at OwnBackup. And a lot of companies that we're working with are in

17:07

the software space,

17:08

but we're also starting to work with companies and other industries as well.

17:11

Because as you can

17:12

see, the technology is broadly applicable, I think, especially when it comes to

17:15

this problem.

17:16

Amazing. Okay. Now, I love this question. What is next on your AI roadmap?

17:21

Where are you guys

17:21

trying to take Triple in the future? Yeah. One of the hard things to crack in

17:26

this space, Sarah, is

17:27

it's a data engineering problem, right? If you look at how much information is

17:33

there within the

17:34

enterprise, everyone is trying to solve a similar problem in the base of AI,

17:38

which is,

17:39

how do we make sense of all this information? How do we structure it correctly

17:42

before we take it in

17:43

in order to make better use of it? A lot of the work that we're doing is really

17:48

being deliberate

17:48

about how we think about homelessness of information, how we provide AI-

17:53

assisted

17:53

tooling in terms of cleaning up data before it gets into the system, and then

17:57

also building out

17:59

ways to measure the ongoing efficacy of the tool and allowing users to really

18:04

see the inner

18:05

workings. Because when you're in this world of handing off a piece of work and

18:10

having an agent go

18:10

complete it, you really want to see the task and what it's doing in the

18:14

intermittent steps,

18:15

because then you have this balancing act of trust versus the output is magical,

18:20

but you do want

18:21

some level of auditing. So all of these are the big questions I think of this

18:24

space, and our roadmap

18:25

really is aligned to, as we release new magical features, how do we make sure

18:28

there's transparency?

18:30

That's amazing. Okay, the very last question is, are there any other AI-powered

18:36

products that

18:36

your team is using in your GoToMarket tech stack that you want to give a shout

18:39

out to?

18:40

Absolutely. There are actually another Salesforce Ventures portfolio company

18:45

shout out to

18:45

Grain. We're using them since day one. You see this, you know, Sunil's note t

18:53

aker or Ray's note

18:53

taker pop up at every meeting that we're in. And what it's done a really good

18:57

job of is just

18:58

help us build this database of all the meetings that we've had. And it's been

19:03

fantastic for

19:04

onboarding new employees, because all of our customer conversations, everything

19:07

that we've done,

19:08

as we think about using them as assets for enablement, it's just been a very

19:13

low maintenance

19:14

way of collecting that information, but then also a high impact onboarding

19:19

asset when we think

19:21

about getting people up to speed on the company. So big shout out to Great. It

19:23

's an awesome product.

19:25

Amazing. Well, Sunil, thank you so much for joining the show. As a Salesforce

19:28

Ventures

19:29

company, I qualified. I love having other Salesforce Ventures and Salesforce

19:32

backed

19:33

and founded companies on the show, because I know you guys can just build

19:36

really cool

19:37

and useful products for your end users. So I love the demo. This has been

19:41

amazing. Thank you

19:42

so much for joining us today. Thank you for having me, Sarah.